For several years researchers in bone genetic diseases have been puzzled by the fact that human osteopetroses presented with a normal or even elevated number of osteoclasts. This was particularly intriguing since in the last 15 years a number of artificial mouse models have become available [1] with defects in the production and differentiation of osteoclasts [2], the defective cell in osteopetrosis. In addition, in other human hematological pathologies, such as primary immunodeficiencies, a defect in differentiation was usually predominant.

Osteopetroses are a heterogeneous group of bone diseases [3–5] in which inherited transmission allowed the first bone researchers to distinguish between a dominant (ADO) and a recessive (ARO) form: this distinction was already settled at the beginning of the last century and was extremely interesting from a clinical point of view. Indeed, the phenotype was very severe in the recessive form, but usually mild or even asymptomatic in the dominant one, which was sometimes dubbed as “benign.”

Retrospectively, the idea that all forms of human osteopetrosis were osteoclast rich was not well founded. Osteoclasts can be demonstrated at autopsy, but there were very few reports of osteopetrosis autopsy [6–9]. The other way to assess the presence of osteoclasts is to perform bone biopsy studies in living patients. However, this is an invasive procedure, albeit modest, and it is rarely performed in infants who are severely ill, as many clinicians believe that is not worth carrying out, since so far it has not contributed anything to patient management. Moreover, due to the abnormal bone remodeling, bone specimens do not always reflect the presence of osteoclasts, since these cells could be present outside of the small area investigated by the biopsy. However, when the RANKL/RANK/TRAF6 axis was identified and mice deficient for these molecules were shown to have an osteopetrotic phenotype, several researchers, including ourselves, tried to investigate these genes in humans, initially without finding any mutations, thus suggesting the idea that such defects did not contribute to human osteopetroses.

Osteoclast-Rich Forms: Genetics and Cell Biology

Osteopetrosis associated to renal tubular acidosis was the first form whose molecular defect was identified [10]. Interestingly, carbonic anhydrase II, the enzyme responsible for this disease, pointed to the acidification process, one of the most pivotal functions of the resorbing osteoclast [11, 12]. In fact, the enzymatic reaction catalyzed by carbonic anhydrase provides the protons which will be extruded by the vacuolar proton pump (VPP) in order to acidify the extracellular microenvironment. Several reviews on the biochemistry of the acidification process have been published and we refer to them for detail [13, 14]. Failure in the vacuolar pump was an obvious candidate in disrupting the resorbing function of the osteoclast, but functional cellular studies were inconclusive. The VPP is a multisubunit complex of more than 10 subunits, among which the a3 subunit was shown to be specific for osteoclasts [15]. The a3 is an isoform of the a subunit of the Vo domain in the VPP and it is encoded by the ATP6i gene, later denominated TCIRG1. Although the expression of this subunit is not as restricted as once thought, since its transcript can be easily amplified from several tissues [16] and has been independently isolated from T lymphocytes [17], its role was intriguing enough to prompt Li and coworkers to inactivate its counterpart in the mouse [18]. The osteopetrotic phenotype shown by this knockout mouse, together with the fact that this gene in humans mapped to chromosome 11 which had been previously associated with human osteopetrosis [19], prompted several researchers to sequence this gene in humans and find mutations in several cases [20–26]. Moreover, the few available bone biopsies, showing that these patients have multinucleated osteoclasts, allowed us and others to conclude that TCIRG1 gene mutations are responsible for an osteoclast-rich ARO [21, 24, 27]. This was also confirmed by Carle and coworkers, who in the same period demonstrated an inactivating deletion in the tcirg1 gene in the spontaneous murine model, called oc/oc [28]. In the meantime, based on the already recognized monocytes ability to differentiate into osteoclasts in coculture experiments with stromal cells [29], reproducible RANKL-driven assays were developed by various researchers [30–33]. Their application to patients [34–36] revealed that monocytes carrying TCIRG1 defects can form multinucleated osteoclasts in vitro, even though these cells are unable to resorb bone, confirming that the VPP plays a fundamental role only in the very late steps of osteoclast physiology.

TCIRG1-Dependent Osteoclast-Rich Form

TCIRG1 is the most frequently mutated gene in human ARO. Indeed, in our cohort of more than 230 ARO patients, mutations in TCIRG1 account for 50% of cases (Fig. 1). All types of mutation are present, with some being found in several patients sharing common ancestry [23, 26, 37]. Curiously, ARO is relatively more frequent in Costa Rica, as has also been reported for other genetic diseases in this country [38], where a founder effect could be hypothesized on the basis of the historical fact that the Indian population was almost completely destroyed and substituted with a small number of Spanish invaders. In particular, we found that two different mutations are widespread in this population, accounting for all the TCIRG1-dependent Costa Rica patients [37], and suggesting that both mutations had been introduced separately by two Spanish individuals, although the possibility of interbreeding between the two populations cannot be excluded.

Fig. 1
figure 1

Distribution of ARO patients in our series, on the basis of the specific molecular defect. The PLEKHM1 slice is too small to appear in the graph since it amounts to 0.5%

Full size image

An additional puzzling aspect regarding the TCIRG1 subset of patients was the finding of a few cases with mutations in only one allele [22, 37], which raised the possibility of a dominantly transmitted disease, although a variable penetrance should be hypothesized, since their heterozygous parents were asymptomatic, similarly to what occurs sometimes with heterozygous ClCN7 mutations. However, recently in four out of fifteen patients, carrying only one affected allele, we detected an internal deletion spanning exons 11–13 in the second allele, strongly suggesting that all these individuals might represent classical biallelic ARO in which one mutation had previously gone undetected. The same genomic deletion was found at the homozygous level in one additional patient [39].

As stated before, monocytes from TCIRG1 patients differentiate when exposed to M-CSF and RANKL but are unable to resorb bone. All the mutations seem to inactivate the protein, and probably for this reason the clinical picture is quite homogeneous, covering all the signs and symptoms of classical osteopetrosis: high bone density, propensity to fractures, no marrow cavity, and pancytopenia with extramedullary hematopoiesis causing hepatosplenomegaly. Of note, neurological defects including blindness and deafness, due to cranial nerve compression, and hydrocephalus, due to skull deformities, are also observed. It is accepted that the neurological deficits are secondary and that the a3 subunit, although expressed in brain, does not have a direct effect on the retinal and nervous cells, probably because other “a” subunits, coded for by different genes, substitute for its function.

ClCN7-Dependent Osteoclast-Rich Form

Extrusion of protons by the VPP is essential for extracellular milieu acidification and a parallel anion pathway is needed to allow bulk proton transport. In addition, it has long been thought that a chloride channel should also be involved in the process to electrically shunt the VPP. There were many candidates for these roles, including proteins of the ClC family which comprises several members [40]. Recent data suggest that some of them, including ClCN7, are antiporters instead of true chloride channels [41]: an evolutionary explanation for the fact that the ClC family contains both true Cl– channels and Cl–/H+ antiporters has recently been proposed [42]. Although their precise biochemical function has only recently been clarified, the genetic approach of ClCN7 gene inactivation in mouse drew attention to its fundamental role in osteoclast physiology, since the knockout mouse surprisingly showed a severe osteopetrosis [43]. The original paper reported a single patient who was compound heterozygote for a nonsense and a missense mutation in the gene, but this observation was further confirmed by Frattini and coworkers, who described seven patients with ClCN7 mutations in both alleles [44]. To date, the frequency of ClCN7 patients in our series is about 15 % (Fig. 1). Since the ClCN7 gene maps to chromosome 16 where van Hul’s group had already mapped a locus for dominant osteopetrosis (ADO-II) [45], the Belgian group was able to show that the ClCN7 gene not only was mutated in ARO but also represented the long-sought gene responsible for Albers-Schönberg disease [46], which is also an osteoclast-rich form [47, 48]. Therefore, the ClCN7 gene covers the complete spectrum of osteopetrotic phenotypes, from asymptomatic high bone mass to early lethal ones. The benign and mild types are always due to a single allele mutations, while severe cases are usually due to biallelic mutations. Intermediate cases, representing an ill-defined entity, are also caused by heterozygous ClCN7 mutations [44, 49], although not exclusively (A. Villa, unpublished). There is, to our knowledge, a single molecularly proven ClCN7 patient on whom osteoclast analysis has been performed, showing their presence at bone biopsy and their ability to differentiate to multinucleated nonfunctional osteoclasts [34]; in another patient bone biopsy was performed, but the presence/absence of osteoclasts was not reported [50]. When ClCN7-recessive patients were compared to TCIRG1 ones, it became clear that the former have more severe involvement of the nervous system [44], probably due to a specific function of the ClCN7 gene in the brain and retinal cells. This has been ascribed to a defect at the lysosomal level [51]. Therefore, as initially suggested by the knockout mouse study [43], the ClCN7-dependent recessive form is an osteoclast-rich ARO with specific primary involvement of the nervous system.

OSTM1-Dependent, Osteoclast-Rich Form

Bone genetic diseases are a field in which murine studies of both artificial and spontaneous mutations have been extremely important [1]. Several spontaneous mouse mutants have been described in the last century, including the gl/gl [52]. The isolation of the gl gene in the mouse mutant strain by Vacher’s group has allowed the identification of a distinct group of patients with mutations in this gene (now denominated OSTM1). So far, seven patients have been reported [53–58]. Seven additional patients are present in our cohort, thus leading to an estimated frequency of about 5% (Fig. 1).

Unfortunately, OSTM1-dependent ARO is associated with a very severe prognosis, due to the presence of severe brain abnormalities and epilepsy [54]. These abnormalities went undetected in the original mouse brain, but we were able to show that they are present also in the mouse [54]. Although the number of reported patients is small, these features are somehow reminiscent of the CLCN7 phenotype, which also presents a primary nervous system defect [44, 51]. Interestingly, a recent report suggests that ClCN7 and OSTM1 proteins can interact, thus explaining their similar presentation [59]. Indeed, a few OSTM1 patients, when initially examined by a neurologist, are suspected to have a lysosomal storage disease [60]. Interestingly, the OMIM database lists an “Osteopetrosis and infantile neuroaxonal dystrophy” (MIM 600329) which could indeed represent an OSTM1-dependent osteopetrosis, although so far none of the original cases have been investigated for mutations in this gene.

A case has been described at autopsy, but the examination was performed before knowing the molecular diagnosis [55]. Osteoclasts were described to be present, but in reduced numbers and with marked morphological alterations, however, detailed studies were not performed. In a recently reported patient, monocytes were able to form mature multinucleated osteoclasts when exposed in vitro to M-CSF/RANKL [57]. These scant data, together with studies on gl/gl mouse strain, suggest that this is also an osteoclast-rich form.

PLEKHM1-Dependent, Osteoclast-Rich Form

The incisor absent (ia) rat, first described in 1941, is a spontaneous rat mutation in which absence of incisors is associated with osteopetrosis [61]. Van Hul’s group has recently identified the gene responsible for this mutant as PLEKHM1, a gene not previously involved in bone physiology [62]. Its protein is suggested to function in vesicular transport since it colocalizes with Rab7 to late endosomal/lysosomal vesicles in osteoclast-like cells. Lysosomes play a major role in osteoclasts, since they have to process all the degradation material taken up from the resorption lacuna; interestingly, the ClCN7/OSTM1 complex also localizes to lysosomes [59].

The original paper described one consanguineous family in which two siblings showed a biallelic homozygous mutation in the PLEKHM1, while a third was heterozygous for the mutation. The osteopetrotic picture of the proband is not as severe as that of TCIRG1-dependent patients, in agreement with the picture of the ia rat, but is definitively recessive in nature and therefore can be classified among the ARO. Her younger brother, who was homozygote for the mutation, was still asymptomatic when the molecular analysis was performed; however, he showed dense metaphyseal bands at 8 months of age and his osteoclasts were defective in an in vitro assay of bone resorption. Although no bone biopsy from these two patients was available, the finding that osteoclasts can differentiate in vitro but are nonfunctional in resorption, together with the observation that ia rats have two to three times more osteoclasts than their normal littermates, allows classification of this disease as an intermediate, osteoclast-rich ARO. So far, only an additional patient with a heterozygous missense mutation in the PLEKHM1 gene, whose monocytes were able to differentiate into multinucleated cells, has been described [63]. These data show that the PLEKHM1 form is the rarest among the ARO, with an overall frequency of <1% (Fig. 1).

CA-II-Dependent, Osteoclast-Rich Form

The carbonic anhydrase (CA-II)-dependent form, the first human osteopetrosis whose molecular defect was identified, is easily diagnosed since the bone phenotype is associated with renal tubular acidosis. A deficiency in this enzyme was first suspected on the basis of enzymatic studies and later shown directly at the gene level [10, 64]. This is also an osteoclast-rich form, since osteoclasts are present, although they fail to produce ruffled-borders [65]. In addition, in the CA-II knockout mouse model the osteoclast number is increased, although a prominent osteopetrosis is absent [66].

NEMO-Dependent, X-Linked Osteoclast-Rich Form

Two male patients affected by X-linked osteopetrosis associated with anhidrotic ectodermal dysplasia (EDA), immunodeficiency and lymphedema (OL-EDA-ID, MIM 300301) have also been described [67–69]. In both cases, the molecular pathogenesis is due to a point mutation affecting the stop codon (X420W) of the NF-κB essential modulator (NEMO) gene, causing the translation of 27 additional amino acids. This mutation is clearly hypomorphic and it is likely that the residual activity is responsible for the milder overall clinical picture which allowed the survival of the two male patients (null mutations are thought to be lethal in males). A third patient bearing a different NEMO mutation (1167_1168insC), with osteopetrosis and EDA-ID but not lymphedema, has also been reported [70]. Bone biopsy showed osteoclast presence in one of the two patients with the X420W mutation [68], but not in the child with the insertion [70]. However, the interpretation of the latter finding is unclear, since the same mutation has also been reported in an EDA-ID patient without osteopetrosis [67].

Osteoclast-Poor Osteopetrosis: Genetics and Cell Biology

Taken together, all the osteoclast-rich forms account for about 70% of cases (Fig. 1). Although all the five genes initially found coded for effector molecules related to the specific resorbing activity of the osteoclasts, a few clues to the existence of other forms in which molecules acting in the differentiation steps were involved, slowly accumulated.

The first observation, though rarely reported [34, 35, 71], was that some patients did not show any osteoclasts in their bone biopsies. Although, as noted above, this cannot be taken as absolute proof of osteoclast absence in any bone region, the second observation was that monocytes from these patients failed to differentiate into multinucleated osteoclasts, corroborating the histological data. It must be kept in mind that the differentiation assay is available only for a small subset of patients for technical reasons, including the difficulties in obtaining material from small, severely ill patients and in transferring fresh blood to the few laboratories which are able to reproducibly perform this test.

Taken together, all these reports, as well as unpublished data, raised the possibility that human osteopetrosis could be due to a defect in osteoclast differentiation. If this were true, there was no shortage of candidate genes. A short list could include RANKL, RANK, TRAF6, FOS, JUN, NFATc1, NF-κB subunits, RGS10, etc., all of which can cause severe, osteoclast-poor osteopetrosis in the mouse [1, 72, 73]. As mentioned above, some of these genes were tested in the nineties, before the identification of the acidification step as the most frequently affected molecular pathway in osteoclast physiology, so if they were responsible for a minimal portion of ARO cases, a mutation could have been missed. Other genes, such as c-fms (the receptor for M-CSF), DAP12, and TREM2, have been shown to cause either osteopetrosis associated with other defects in mouse [74] or specific bone diseases different from pure osteopetrosis in humans [75, 76].

RANKL-Dependent, Osteoclast-Poor Form

RANKL (TRANCE, OPGL, TNFSF11) is the master cytokine driving osteoclast differentiation in humans and mice [77]. RANKL knockout mice have long been known to display both osteopetrosis and immune defects [78, 79], among which the most obvious is the absence of lymph nodes. The osteopetrosis observed is clearly of the osteoclast-poor type since differentiation cannot even be started in absence of RANKL. Although an in vitro RANKL-independent pathway has been described [80–82], this alternative route appears to be of modest relevance in vivo since osteopetrosis is severe in rankl −/− mice. The defect in rankl −/− mice is noncell-autonomous, since RANKL is secreted by stromal cells, including osteoblasts, and is unique because monocytes from rankl −/− mice can be induced to differentiate normally into multinucleated functional osteoclasts in the presence of M-CSF/RANKL, unlike monocytes from mice with deficiency in genes downstream of M-CSF or RANKL signaling. Therefore normal osteoclast differentiation in the presence of RANKL can be used to infer a RANKL defect in ARO patients.

After selecting from our large series a few patients whose bone biopsy was reported to be negative for the presence of osteoclasts (necessary, but not sufficient to prove an osteoclast-poor origin of the disease) and who were negative for mutations in the five genes responsible for osteoclast-rich ARO, we sequenced RANKL and found that four of them had mutations in this gene [83]. To further confirm the involvement of RANKL gene in these patients, we performed the in vitro differentiation assay on monocytes from two of them who had not undergone transplantation, and we found that they were indeed able to form resorbing multinucleated osteoclasts, as predicted. Interestingly, purified hematopoietic stem cell (HSC) transplant, performed in three patients was only partially effective, since, although some hematological parameters were improved, the bone defect was not rescued, in agreement with the fact that RANKL is produced mainly by mesenchymal cells, which were not transplanted in our cases [83]. We concluded that RANKL-dependent osteopetrotic patients truly represent an osteoclast-poor ARO subset.

RANK-Dependent, Osteoclast-Poor ARO

Patients without osteoclasts in bone biopsies who did not show mutations in the RANKL gene were sequenced for other genes, including RANK. Indeed, mutations were detected in some of them as well as in other patients for whom no bone biopsy was available. A total of eight patients from seven unrelated families were found to bear RANK mutations [84]. Interestingly, mutations (in frame duplications) in the RANK signal peptide were described several years ago, but the phenotype of these patients is clearly different [85]. They have been classified under the diagnosis of early-onset Paget disease (MIM 602080, PDB2), familial expansile osteolysis (MIM 174810, FEO) or expansile skeletal hyperphosphatasia (ESH), in relation to the exact number of triplets contained in the duplication. Clinically, these patients are distinguished from osteopetrosis patients by their accelerated bone turnover detectable by biochemical markers, by the focal nature of the bone involvement, and by the dominant inheritance of the condition [85–87].

Rankl- and rank-defective mice show very similar phenotypes, with osteopetrosis and lack of lymph nodes. However they differ with respect to their ability to differentiate osteoclast precursors in vitro after exposure to M-CSF/RANKL since in rank-deficient mice the defect is cell-autonomous. This means that while RANKL-deficient patients produce functional osteoclasts, RANK-deficient ones are expected not to do so. In agreement with this, when cells from RANK-deficient patients were forced to differentiate in vitro, no multinucleated osteoclast was seen. In addition, transplanted patients showed a rescue of the bone phenotype [84]. We concluded that the RANK-defective subgroup represents a second type of osteoclast-poor ARO and that the response of osteoclast precursors to growth factor stimulation in vitro and of the patient to HSC transplant in vivo can faithfully discriminate between the two subsets of patients.

Osteoimmunology

Osteoimmunology is a term recently coined to indicate a new field of study [88]. Since the description of abnormalities in both the skeletal and the immune system in RANKL, RANK, and TRAF6 knockout mice, it has become clear that complex interactions underlie the relationship between the two systems and that researchers involved in the study of both fields were fated to meet and share their knowledge. The interplay between the two systems has proved to be wider and wider, with the identification of several other molecules which lead to defects both in bone remodeling and the immune response, when mutated [89, 90].

Interactions between immune cells and osteoclasts were originally shown in murine models by gene targeting of molecules involved in the early steps of osteoclast differentiation; subsequently, similar observations were reported in humans [91, 92].

We performed immunological investigations in two subsets of ARO patients bearing mutations in RANKL and RANK. Our results can be regarded as preliminary, because of the limited blood amount available from these young and severely ill patients, which limits the number of assays that can be performed.

Among six RANKL-dependent cases, we analyzed two untransplanted and two transplanted patients, as controls, and revealed a slight immunological defect in T cell function, namely, a partial defect in proliferation and Th1 and Th2 cytokine production [83], together with a lack of palpable lymph nodes. This mild deficit in patients is clearly in contrast with the more severe phenotype reported in the corresponding mouse model, where a block in T cell differentiation at the pre-TCR stage and B cell differentiation at the proB-preB stage, together with a complete absence of lymph nodes, has been described [78, 79].

Regarding RANK-deficient patients, immunological investigation was performed only in untransplanted patients. In particular, two affected siblings suggested that the RANK mutation is associated with a partial defect in peripheral B cell maturation, since a reduced number of memory switched cells was found. Moreover, immunoglobulin serum levels, assessed in four out of eight patients, were reduced in three of them. In addition, three patients, failed to respond to a full course of tetanus toxoid vaccination. As for the RANKL-dependent subset of patients, the defect described in the RANK-dependent ones is milder than in the murine counterpart, which strictly resembles the rankl −/− mouse in its immunological phenotype. In both cases, the discrepancies could be due to species-specific defects, as already reported for other bone diseases [93, 94]. However, additional studies and a more comprehensive analysis involving other immune cell types and, possibly, other patients would be required to test this hypothesis. Of note, the three NEMO-dependent patients showed low Ig levels similar to those reported in RANK-defective ARO [67, 70]. What is clear now is that clinicians and researchers can no longer focus on a single system or tissue, but they have to take into account the interactions between them.

ARO Heterogeneity, Prognosis, and Indication for Bone Marrow Transplant

Molecular diagnosis is able to dissect the heterogeneity of ARO and study of genotype/phenotype associations which may be of clinical relevance (Table 1). A schematic representation of the genes responsible for human ARO in relation to their function in the physiology of the osteoclast is depicted in Fig. 2, showing that what was once a relatively homogeneous “malignant infantile autosomal recessive osteopetrosis” can now be classified in several subsets with some peculiar clinical aspects.

Table 1 Different forms of osteopetrosis in relation to the molecular defect and the possible therapy
Full size table
Fig. 2
figure 2

Schematic representation of differentiation from hematopoietic stem cells to functional, resorbing, multinucleated osteoclasts. Genes known to be involved in the pathogenesis of human ARO are underlined. RL resorption lacuna, SZ sealing zone. Bottom, left: confocal image of an osteoclast cell differentiated in vitro, on glass, from PBMCs of a normal donor and histochemically stained for vitronectin receptor αvβ3 (red) and actin (green); nuclei are stained with DAPI (blue). Scale bar: 100 μm (refer color images in online)

Full size image

The first consistent feature is that two osteoclast-rich forms, ClCN7- and OSTM1-dependent ARO, present with severe nervous system impairment that makes their prognosis very poor [44, 54, 55, 57]. The involvement of the nervous system in some cases is so predominant that a diagnosis of lysosomal storage disease may be suspected, especially when skeletal X-rays are not performed early. This is due to the role that ClCN7 and OSTM1 play in the brain [51, 59], and although only a genetic test can definitely definitively confirm the diagnosis, an OSTM1- or ClCN7-dependent form can be suspected on simple clinical grounds. It is now agreed that bone marrow transplantation (BMT) is not indicated in the former subgroup, while in the latter one, BMT must be considered in relation to the severity of nervous system involvement [44, 54].

A TCIRG1-deficient form usually presents with classical osteopetrosis and with secondary nervous system involvement. In the presence of severe neurological signs, cranial foramina neurosurgical decompression can be useful. The TCIRG1 form is clinically quite similar to the osteoclast-poor RANK-deficient form. Indeed, both subgroups benefit from HSC transplantation, which in both forms can cure the bone phenotype, although cranial nerve compression and growth retardation often do not improve.

The PLEKHM1 form is recessive and extremely rare and has a phenotype of intermediate severity which can be easily separated from other ARO forms, as well as from the ADO on radiological grounds. The only two patients described so far have not been transplanted due to their mild clinical phenotype [62, 63], and it is likely that for this reason BMT would not be appropriate to treat these patients.

Although only a few RANKL-dependent patients have been described so far, it seems that they have a slightly less severe course since the majority of nontransplanted patients are still alive [83]. If a few osteoclasts form along a minor RANKL-independent osteoclast differentiation pathway in vivo, as shown to occur in vitro [80, 81], these cells could display the full pattern of bone matrix degrading enzymes, allowing a small degree of resorption. Classical BMT, which is often performed with purified HSCs, can restore the hematological defect, but not the bone phenotype, since mesenchymal cells are not transplanted in this case. Indeed this is what occurred in our patients who were treated with BMT before their molecular diagnosis was defined. Further studies will show whether a population of mesenchymal stem cells can be enriched and transferred together with HSC and whether this would be sufficient to rescue RANKL production.

Future Therapies

Bone marrow transplantation is the only therapeutic approach available in ARO management [95, 96]. The success of BMT depends on the engraftment and graft-versus-host disease (GVHD) absence, which in turn relies strongly on major histocompatibility complex (MHC) matching [97]. In addition, despite successful engraftment, the beneficial effects are strictly dependent on the time of the transplant [97, 98]. ARO skeletal deformities start before birth and they are not easily reversed, particularly in the skull where narrow foramina can entrap cranial nerves. Growth is also affected and is not cured by the transplant. Therefore ARO is a paradigm for all those diseases whose stigmata are already present at birth and cannot be reversed by postnatal treatment.

Several studies have tried to investigate whether early BMT could improve the quality of life of these patients by using the mouse model of TCIRG1-dependent ARO, the oc/oc mouse. The underlying molecular defect in this strain is a homozygous mutation inactivating the a3 gene, leading to a “null” phenotype which closely resembles human ARO, making the use of this mouse relevant to the human situation [28]. Despite an old report suggesting that the oc/oc mouse was resistant to BMT [99], further experiments in both humans and mice have definitively shown that TCIRG1-dependent ARO is sensitive to BMT. Johansson and coworkers took the approach of performing neonatal (1-day-old) cell therapy either with or without additional gene therapy based on viral vectors [100, 101]. They obtained encouraging results, with normalization of several parameters, although the mice remained smaller than controls. However, it is unlikely that such early treatment could be performed in humans, since it usually takes a few months for these patients to be diagnosed, unless the family had already had an affected sibling; in addition, the authors used MHC matched cells, which are not always available in humans. Of note, Frattini and coworkers proposed an alternative approach, in utero BMT with adult cells [102], a strategy already used in humans, especially to treat patients with primary immunodeficiencies [103]. In these experiments, to better simulate the human situation, unmatched cells were used. As expected, unmatched donor hematopoietic cells were easily tolerated and no GVHD phenomenon was detected. Complete rescue of the phenotype was achieved in a portion of the treated mice, with the exception of tooth eruption, which seems to be the most difficult defect to correct [100, 102]. Although the percentage of totally cured mice was not high, an improvement was detected in most of them, and there is the possibility that the use of other types of hematopoietic stem cells (for example, fetal liver or cord blood cells) could increase the cure rate. Obviously, clinical application is so far very limited, since the in utero approach is restricted to families planning a further pregnancy after the birth of an affected child. However, the availability of molecular prenatal diagnosis is becoming more common, paving the way for new therapeutic strategies.

The detection of a subset of RANKL-dependent patients opens the road to new therapeutic opportunities: while these affected individuals do not respond to HSC transplant, they could benefit from mesenchymal stem cell transplantation (MSCT). This kind of approach, even though based on a cell type which is still quite elusive and difficult to unequivocally identify, definitely has considerable potential. MSC transplants are currently under consideration in several fields [104–107]. With respect to RANKL-dependent ARO, the rationale in the application of MSCT is that in this way the patient would receive the specific precursor cells able to differentiate into normal osteoblasts, which in turn would produce in situ the cytokine (RANKL) originally defective in the recipient, thus ultimately allowing the differentiation of osteoclast precursors into normal, functional cells. Besides the possibility of exploiting cellular approaches other than HSCT, the identification of RANKL-dependent ARO paves the way also for pharmacological approaches, as, at least in theory, these patients could benefit from the administration of recombinant RANKL. For this purpose studies in animal models are required, and if positive results supporting the feasibility of this approach also in humans are obtained, major efforts will be justified in order to make biotechnological firms meet the world of “orphan” diseases.